pickle-mixin

Makes un-pickle-able objects pick-able.

Install

You can install it via pip

pip install pickle-mixin

Running the tests

After installation, you can test it

python -c "import pickle_mixin; pickle_mixin.test()"

as long as you have pytest.

Usage

Pickle by initialisation

Suppose that you have a class whose objects are un-pickle-able or that would demand a large amount of disk space or memory to be pickle-able. PickleByInit class lets you pickle object attributes via object initialization. Consider the following classes:

class Foo(PickleByInit):
    def __init__(self, obj):
        super(Foo, self).__init__()
        self.obj = obj

class Bar(object):
    def __init__(self, filename):
        self.filename = filename

    def __getstate__(self):
        raise PicklingError

    def init_dict(self):
        return dict(filename=self.filename)

Trying to pickle as follows

f = Foo(Bar('file.txt'))
pickle.dumps(f)

would raise a PicklingError. The following on the other hand would work:

f = Foo(Bar('file.txt'))
f.set_signature_only_attr('obj')
pickle.dumps(f)

The un-pickling process of f.obj attribute happens via object initialisation, passing the returned dictionary from init_dict as keyword arguments to Bar.__init__.

Mixing classes with and without slots

Pickling does not save attributes defined via __slots__ in the following case:

class Foo(object):
    __slots__ = ['a']

    def __init__(self):
        self.a = 4


class Bar(Foo):
    def __init__(self):
        pass

SlotPickleMixin fixes it:

class FooMixin(object):
    __slots__ = ['a']

    def __init__(self):
        self.a = 4


class BarMixin(FooMixin, SlotPickleMixin):
    def __init__(self):
        FooMixin.__init__(self)
        SlotPickleMixin.__init__(self)

f = BarMixin()
o = pickle.dumps(f)
f = pickle.loads(o)
assert hasattr(f, 'a')

Authors

• Danilo Horta - https://github.com/Horta

License

This project is licensed under the MIT License - see the LICENSE file for details